Welcome to Apex Tube

Apex Tube Company is a typical discrete parts manufacturer that we will use to
illustrate the process of creating continuous flow. Apex produces a variety of tubular
products for automotive, truck, and heavy-equipment applications. Two years ago
Apex responded to pressure from its customers for lower prices, higher quality,
more frequent deliveries, and more rapid response to changing demand by taking
a hard look at its manufacturing operations.

For many years the company had organized its fabrication and assembly processes by
department with each product visiting each department as necessary. The resulting
maze of spaghetti-like product movements was hard to manage and even harder to
improve. Apex managers therefore took the first step recommended in Learning to
See and conducted an analysis of their products to find product families that could
be managed individually.

Apex managers drew up a product family matrix that grouped products by similar
sequence of final processing (pacemaker) steps and machines.

Apex’s Product Family Matrix

Assembly Steps and Machines

end sub- final
form pierce braze bend crimp test
assembly assembly

automotive
X X X X X
X X X X X X
PRODUCTS

truck S

truck L
X X X X X X
truck A
X X X X X X
heavy truck

heavy
X X X X
equipment
X X X X X X

PART I: GETTING STARTED 1

 The light-truck product family made the greatest revenue contribution to Apex and was
under the heaviest price pressure. Apex appointed a Value Stream Manager for this product
family, who drew a current state value stream map. This product family is shipped to the
State Street assembly plant in three variants: a short-hose assembly (S) for the short
wheelbase truck, long-hose (L) for the long wheelbase model, and an alternative-fuel
(ethanol) assembly (A) offered as an option on this vehicle.

PRODUCTION
Forecast CONTROL

MRP
Michigan Weekly
Steel Co. Order

Weekly Schedules

Tues. &
Thurs.

TUBE EXTRUSION END FORMING TUBE BENDING SUB-ASSEMBLY

Coils 5520 2760 2484

5 days 1 1 3726 1 1

C/T = 7 second C/T = 12 seconds C/T = 24 seconds C/T = 22 seconds

C/O = 1 hour C/O = 10 minutes C/O = Ø C/O = Ø

5 days 4 days 2.7 days 2 days 1.8 days

7 seconds 12 seconds 24 seconds 22 seconds

2
 With their batch and queue production system based on a process village layout, Apex
managers were not too surprised to learn that actual processing time was less than 0.01%
of lead time and that much of the floor area devoted to this product was either for storing
inventory or to provide access between process steps. In short, nothing flowed and it was
very difficult for Apex to respond to changing customer requirements even with very large
in-process inventories.

Forecast

State Street
Daily Assembly
Order

1,380/day
Box = 30 pieces
2 Shifts
Apex Truck Fuel Lines
Original State Map

4x
Daily Ship Daily
Schedule

FINAL ASSEMBLY
CRIMPING TEST
SHIPPING

Staging
2 1 1
3174 2760 4140
C/T = 37 seconds C/T = 20 seconds C/T = 19 seconds
C/O = 20 seconds C/O = Ø C/O = Ø

Production
2.3 days Lead Time = 22.8 days
2 days 3 days
74 seconds 20 seconds 19 seconds
Processing
Time = 178 sec.

PART I: GETTING STARTED 3

 Apex managers quickly decided to create a continuous flow cell for the five final fuel line
processing steps. (Before doing this they made sure, of course, that enough machines were
still available in the process villages to sustain production for the other products in the plant.)
Shortly afterward they also developed supermarket pull systems between the new cell and
the two upstream fabrication steps that would continue to operate in a batch mode shared

Production Control
Forecast

Michigan Steel
Co. 2 x Weekly
Orders

Tues
& Thurs
il
co

Tube Extrusion End Forming

C/T = 7 seconds C/T = 12 seconds

C/O = 1 hour C/O = 10 minutes

3 days 3 days
7 seconds 12 seconds

4
 across several product families. These pull systems replaced the schedules previously
used to regulate these processes. Through hard work and by suspending traditional
rules of thumb on how quickly change could happen, Apex was able to design and
implement the future state shown here.

Forecast State Street

Assembly
Daily
Order
1380/day
Box = 30 pcs.
2 shifts
Apex Truck Fuel Lines
Future State Map

Daily Ship Schedule

4x
Daily

30
30
30

Assembly Cell S Shipping

Staging
L

A
4 Operators
C/O = 20 seconds

Production
Lead Time = 10 days
2 days 2 days
159 seconds
Processing = 178 sec.
Time

PART I: GETTING STARTED 5

 Apex started its fuel line value stream improvement at the right place: the ‘pacemaker’
process. The pacemaker involves production steps that are dedicated to a particular family
of products and responds to orders from external customers. A well run pacemaker sends
smooth demand signals upstream to the pull loops of the remaining batch fabrication
processes, which respond to requirements from internal customers.

Apex managers and engineers made another good decision by minimizing their initial
investment and keeping the cell simple. For example, they could have created a more
extended continuous flow by incorporating an end-forming press into the fuel line cell.
But such a press would have required substantial capital investment. (In the future Apex
may decide to apply some of the cost savings from its improvements to purchase and add
a press to the cell.) They decided to install a simple, inexpensive, flexible operator-based
cell designed for State Street Assembly’s needs. This is more likely to be highly reliable
and well-suited to sending smooth signals up the value stream.

Apex managers chose a classic U-shaped layout for their new operator-based cell, as illustrated
on the next page. In only a few days they were able to move machines and configure the
new cell to achieve striking reductions at this process in lead time and floor space required,
while dramatically increasing the number of pieces produced per production associate.

Original Current
State State

Apex’s
Continuous
Progress with Flow
No No
Continuous Flow
unstable unstable
Production
per Shift ≈622 ≈622
(actual/target) 690 690

Space 1130 580

(sq. feet)

Assembly 11 days 37 min

Lead Time
(WIP x Takt)

Number of 6 4
Operators

Productivity 13.5 20
(pieces/associate/hr)

Functions No No
Effectively as
Pacemaker

6
 A Closer Look — With Eyes for Flow
Apex managers, engineers, and production associates were excited about their new
fuel line cell. After all, they quickly increased productivity by 50% while halving
space requirements and dramatically slashing lead times. Yet when you look at the
Apex cell with “eyes for flow” you should actually be disappointed. A walk through
the fuel line cell will show why.

Apex’s fuel line cell — current state

material flow

Assembly II
I
bly

Cr
7 pcs.

im
m

3 pcs.
se

pe
As

r
2
1 3

12 pcs. 9 assemblies
(to be tested)
Tube Bender Tester
(automatic) out (automatic)
25 pcs.
Hourly Production
load 25 tubes
load
4
at a time Plan Actual
1 90 65
2 90 79
400 Tubes packout
3 75 80
(raw materials) 30 pcs/container
4 90 92 (finished goods)
5 90 76
6 90 63
7 75 69
8 90 98
690 622

PART I: GETTING STARTED 7

 The first step when we visit a facility is typically to go see the current situation with
our own eyes and ask, “What is the problem?” At Apex, the first thing we notice is
the production output chart at the entrance/exit of the cell showing planned and
actual production. Looking at the output figures we wonder, “Why is there so much
variation, and why does total production fall short of planned production?” More
specifically, “Why is the cell achieving only two-thirds of planned output during
many hours of the shift?” Is the problem incapable machines that make bad parts?
Is it machines that won’t run? Is a supplier shipping bad parts, or are parts missing?
And who reacts when these problems occur?

Whatever the cause, the variation in output is clear evidence that cell performance
can be greatly improved. We are even more certain of this when we note that in two
hours out of eight the cell actually produced more than the plan, which is just as
bad as being under the target. Four production associates were assigned to the cell
the entire shift, so a change in staffing can’t explain the variations. Unless this was
achieved by hurrying, unacceptably risking stress injuries and bad quality, there
must be waste in the process.

We begin to see a source of variation and waste when we closely examine the first two
steps in the production sequence: the tube bender and the first assembly operation.
The first production associate has to leave her regular work area every 25 pieces, or
about every 16 minutes if the cell is producing to takt time (as explained below in
Question 2). This requires three minutes and means that either the material flow
stops or the tube bender and the first two assembly steps are all decoupled from
one another. This means no continuous flow.

As we continue to walk around we notice that there are various quantities of inventory
between every operation and that the production associates are each anchored to their
machine, which often means they have to wait while the machines cycle.

Variable inventory buffers between workstations are an inefficient way to balance

uneven workloads. When a buffer gets too full, the supplying operation often takes
an unofficial break — perhaps to get materials or do other out-of-cycle work — while
the downstream station catches up. Operations are decoupled, allowing each to
produce batches instead of one piece at a time.

8
Decoupled operations, which we call ‘islands’, bake the waste of THREE FLOWS
overproduction and the waste of waiting into a cell, causing them
to be repeated many times every shift, day, week, month, and year. As you look at your own
Tiny wastes often don’t seem significant to managers just visiting cells or lines, tune your
the process (and apparently are not visible to Apex managers), but eyes to check...
think about them as they add up more than 600 times per shift!
1. Does information flow?
Decoupled operations also make it difficult to notice production - Does everyone know
problems as they happen. When a problem occurs the rest of the the hourly production
target?
stations keep on working. By the end of a shift the unnoticed
problems add up and the production volume falls short of the target. - How quickly are
Pacemaker processes, in particular, need to be manageable. Problems problems and
or abnormalities need to be spotted as they occur and support abnormalities noticed?
personnel must respond to them quickly. Production associates
- What happens when
cannot react to and fix significant production problems, find and there are problems and
eliminate the causes of those problems, and at the same time still abnormalities?
achieve full production!
2. Does the material flow?
Finally, as we complete our tour we note that the Apex cell is laid
- Does the workpiece
out in a very wide “U”. This defeats one of the main objectives of move from one value-
a U–shaped cell layout: Permitting flexible deployment of operators adding processing step
by moving work areas into close proximity. Both the first and last right to the next value-
production associates are moving back and forth over considerable adding step?
distances to handle materials. Flow stops every time they leave a
station to backtrack. 3. Do the operators flow?
- Is the operators'
Our conclusion, at the end of our walk through Apex’s cell, is that work repeatable and
there is actually no continuous flow anywhere. Instead we see only consistent within
erratic and intermittent flow — as indicated by the small piles of each cycle?
inventory between each machine and the fluctuating output from
- Can the operator
hour to hour. Indeed, this cell is really just a ‘module’ of adjacent
efficiently go from
machines and operators producing at best ‘fake flow’ that misleads performing one value-
the untrained eye. adding work step (work
element) to the next?
 Targets for Apex’s Fuel Line Cell

Original Current
State State Target

Continuous
Flow No No Yes

unstable
Production
per Shift ? ≈622 690
(actual/target) 690 690 690
Space 252
(sq. feet)
1130 580

Assembly
Lead Time 11 days 37 min 200 sec
(WIP x Takt)

Number of 2
Operators 6 4

Productivity 13.5 20 40
(pieces/associate/hr)

Functions
Effectively as No No Yes
Pacemaker

While Apex’s new cell performance is much better than the original process village layout,
a careful effort to achieve true continuous flow through proper process design and operation
can double labor productivity, halve the needed space, reduce lead time by a further 90%,
and dramatically improve both quality and responsiveness to customer requirements.
Realistic targets for this cell, which we will show you how to achieve in the pages ahead,
are shown in the right hand column of the table above.

We’ll get started by posing the first of eleven questions you should go through as you strive
to develop true continuous flow in your own cells and lines. The questions require careful
work and attention by your entire team, but you will discover that the answers are invaluable
once they are incorporated in your business.

10
Question 1: Do You Have the Right End Items?
Apex has already determined their product families and assigned three end items to
its fuel line cell. However, as you consider your own situation, you may have to think
carefully about the right products to assign to your pacemaker process. Here are some
guidelines we’ve found helpful.

1) Flexibility. Sometimes demand is high enough to allow you to dedicate individual

products to their own cells or lines like this:

Product A Product B

However, if demand gyrates between products and you can keep changeover times short,
you are often better off sharing products between mixed-model cells like this:

Products A & B Products A & B

The total capacity is the same in both cases but the ability of each process to accommodate
shifts in demand between the two products is much greater in the second case. The
demand for one product within a family may vary, while the demand for a whole product
family is often more stable.

PART I: GETTING STARTED 11

 2) Variation in Total Work Content. The total work content — that is the operator time
required to process one piece from start to finish — should not vary by more than about
30% between the different end items processed in the cell, especially when a moving
conveyor is used. When the work content varies too much it becomes difficult to maintain
flow and productivity. In such cases you may want to split the cell or assign some rare or
low-volume end items to other cells. (Some facilities even create a separate line or cell to
handle low-volume end items, until product engineers can reduce the content differences
between the items via design changes.)

3) Similarity of Processing Steps and Equipment. When the steps required to build
different products within the cell vary too much (i.e., when some products skip some
processing steps) operators will have to “shift gears” every time they change to assembling
a variant of the product. This reduces productivity and increases the chance of quality
problems. Again, sometimes it is better to produce variants with markedly different
processing steps in different cells.

4) Takt Time (Production Pace). Takt time is the rate at which customers require finished
units. It is determined by dividing the total available production time per shift by the
customer demand rate per shift (see the equation at right). As a general guideline, when
takt time for a cell falls below ten seconds the operators’ jobs may become highly repetitive
and stressful. When high demand calls for very short takt times you should consider using
multiple footprints of the cell, possibly side-by-side, instead of a single high-speed cell.
This is particularly appropriate if the capital requirements of additional cells can be kept
low through utilization of simple equipment.

Conversely, when takt time slows to more than about 120 seconds, the number of work
elements sometimes gets so high that work motions can be difficult to standardize. In such
cases consider adding additional but similar end items to the cell to bring down the takt
time. Of course, with some products it will simply be impossible to set takt times below
120 seconds because volume requirements are inherently low, even when several different
end items are run through the same cell or line. (With long takt times it can get difficult to
have all parts at the line for the operators for the different product variations. Sometimes you
have to increase the parts delivery frequency or deliver certain parts in the assembly sequence.)

5) Customer Location. When customers for a product are widely dispersed geographically,
it may make sense to split up the work into multiple lines, each located near a different
customer. This makes sense particularly when shipping costs and duties for finished units
are high, when there are potential exchange-rate losses, when lead times for components are
long, or when local infrastructure (supervision, buildings, etc.) is available at reasonable cost.

12
Question 2: What is the Takt Time?
Having decided what products to produce in the pacemaker, the next task for Apex managers
was to determine the takt time. (‘Takt’ is a German word for a pace or beat, often likened
to a conductor’s baton.) Takt time is a reference number that is used to help match the rate
of production in a pacemaker process to the rate of sales.

takt time
Used to help synchronize pace of production with the pace of sales

your available work time per shift

takt time =
customer demand per shift
40 sec.

27,600 seconds
example: = 40 seconds
690 pieces

this means: The customer is buying this product

at a rate of one every 40 seconds.

Sales are usually calculated on a daily or weekly basis but most pacemaker processes are
actually up and running only some fraction of each day or week. Since the point of takt
time is to pace actual production, the most sensible thing to do is to divide the number of
products demanded daily or weekly into the number of shifts operated in that time period
to determine demand per production shift. For example, the customer demand for Apex’s
light truck fuel lines is currently 6900 units per week and Apex operates its fuel line cell
ten equal shifts per week. Thus the demand per shift is 690 units.

Once demand per shift is known the final step in the calculation of takt time is to divide
this number into the ‘effective working time’ per shift. This is start-to-stop shift time
minus any scheduled operator breaks, meetings, cleanups, etc. Because takt time must
represent the actual customer demand rate do not subtract time for unplanned machine
downtime, changeovers, or other internal problems.

PART I: GETTING STARTED 13

 Apex operates two 8-hour shifts Monday through Friday, 6:00 AM to 2:30 PM, and 3:30 PM
to Midnight. There are two 10-minute breaks each shift but no scheduled downtime for
maintenance. This means Apex has 27,600 seconds of effective working time in each shift.

480 min. (8 hours) – 20 min. of breaks = 460 min. x 60 sec./min. = 27,600 seconds

By dividing 690 units into 27,600 available seconds we determine the takt time: 40 seconds.

27,600 seconds
= 40 seconds per unit
690 units

This is the rate of customer demand, the all important ‘beat’ of the market. Notice that
takt time is expressed in ‘seconds-per-unit’ because it is easier for everyone to understand
and use than decimals of minutes. Similarly, we use ‘seconds-per-unit’ rather than ‘pieces-
per-hour’ to describe actual production rates, or ‘cycle time’. Comparing takt time and cycle
time is the easiest way to answer the simple but critical questions: “How frequently does
the customer need one piece?” and “How frequently do we actually make one piece at our
pacemaker process?”

There is one additional point that may be very important in your own takt time calculations,
the amount of variation in customer orders. In Apex’s case the 6900 unit per week demand
was relatively easy to determine because Apex is supplying a massive automotive assembly
plant whose own takt time does not change frequently. But what if long-term average
demand and day-to-day actual demand are different?

We suggest that you check the range of daily customer demand variation by reviewing
actual shipments (not orders) over the past twelve months. Your cell must be able to handle
sustained demand. For occasional spikes in demand it is generally better to operate at a
steady takt time (based on average long-term demand) and either hold a buffer stock of
finished goods or run some daily overtime to ensure your ability to serve the customer.
Changing takt time from day to day is inefficient, disrupts the work pace, and increases
the potential for quality problems.

Lastly, regarding future demand for new products, it can be difficult to make accurate
forecasts far in advance. When future demand is uncertain it may be wiser to add capacity
in steps, as increased demand actually materializes, rather than designing your pacemaker
now for a peak demand that may not appear.

14
Cycle Time
Cycle time is how frequently a finished unit actually comes off the end of your pacemaker
cell. We often find processes that are operated at cycle times faster than takt time. For
example, if you are running your facility three full shifts (perhaps to achieve high machine
utilization) you will probably always need cycle times slightly below takt time because
there is never any time available to catch up if your equipment or materials system fails.
And to some degree these sorts of problems will always occur in manufacturing!

However, keep in mind that when you chronically cycle much faster than takt time you
increase the chances of overproducing and may be using extra operators. (As the diagram
below shows.) Much worse, you conceal your production problems and reduce the incentive
to find and eliminate their causes. It is important to maintain a certain tautness at the pacemaker
to ensure that problems get noticed quickly and receive fast response by support staff.

Cycling much faster than takt may require more people

• • • • • • • • • •

takt time

cycle time
M6

extra operator

Note:
The inevitability of problems in manufacturing is one of the reasons why many
production facilities in the Toyota group of companies run their pacemaker
processes for two shifts with a one to four hour gap between shifts. Then there is
time to make up production losses with a little overtime at the end of each shift.

PART I: GETTING STARTED 15

 Setting the Pace
As you go through the calculations to determine your own pacemaker takt time we need
to explain one final point: It is seldom the case that there is only one correct takt time!

Remember that takt time is customer demand (which you can’t change)
divided into available production time (which you can change.)
12 Specifically, you can adjust:
3
1) The available production time — the number or length of shifts.
6 2) The number of end items produced in a cell.
3) The number of cells making a particular end item.

The pace of production is one of the most critical considerations for the design of your
processes. Here you will often have some choices to make. For example:

• A cell that has a takt time of 40 seconds over two shifts could also be run at 20-second
takt in only one shift. In some cases it is easier and less costly to manage only one shift,
particularly if running a second shift means extra support structure and paying night
premiums. An added bonus is that the waste of waiting time is easier to see and
eliminate when takt time is shorter.

• The size, weight, and complexity of a product can influence what is a reasonable cycle
time and the number of motions for each operator. Producing a light, low-complexity
product with only a few work elements per operator to a 10-second takt time may be fine.
But when operators are working on larger, heavier or more complex products it can be
better to work to a longer takt time and assign more work elements to each operator.

• When new products are introduced, substantial savings in capital investment can be
achieved by adding them to existing cells rather than building additional cells. This will
decrease the takt time for those cells.

• As you launch your new cell it is often much better to utilize a temporary and separately
held ‘safety stock’ of specific finished goods to protect your customer and to set your
cycle time only slightly faster than takt time. The tension this produces forces you and
your staff to address the causes of production interruptions.

With experience you will gradually learn what’s best for you. The key point for the moment
is that you must know what the takt time is and how it was determined.

16